Now that generative AI services like ChatGPT are firmly entrenched in technical workflows, technical professionals with the right knowledge base can take on new jobs. Companies building and researching AI are looking for experts in various AI and machine learning (ML) subfields. Even companies that don’t specialize in AI are creating in-house AI tools for unique functions (such as searching internal databases).
Let’s take a look at what these new jobs are, what they require, and what background they require. And who knows, maybe one of them will jump out at you as your next career choice.
Ethics experts and AI legal experts
Ethics has always played a big role in computers (cookie tracking comes to mind), but AI brings with it a huge amount of new concerns and concerns. Here are some random concerns that companies building AI need to consider.
- If someone using an AI provides personal information within a prompt, will the AI be allowed to save and add those details to the knowledge base?
- Furthermore, does the situation change if such personal information is “anonymized”?
- AI may not be able to distinguish between truth and lies. What happens if an AI spreads lies about a political candidate? Can that candidate sue if they lose the election?
- It is said that someone asked the AI how to make dangerous substances used in chemical warfare. AI refused. The person then allegedly asked the same AI what chemicals should not be mixed together to make the substance, and the AI explained in detail what should not be done. Can AI be held responsible for indirectly instructing how to make dangerous substances?
- Can AI give advice? Although this example has not been confirmed, there is currently an article circulating on the web claiming that Google’s AI told them that someone on Reddit had suggested a “cure” for depression that was pretty unsavory and cruel. Should AI do that? And if people follow such bad advice, could AI companies be sued?
- When an AI that generates images “learns” and examines works by contemporary artists (i.e., works that are still copyrighted), can it use what it learns from those images to construct images that look completely different but have the same general “look and feel” as the original? Can it do that? Should AI get permission? Should AI pay original artists?
Some of these questions can also raise serious legal issues. Courts will issue decisions and lawyers will need to learn about these new precedents. And soon there will be lawyers specializing in AI. Downstream, this means that AI-focused technical professionals will also need to learn as much as possible about ethics and even specialize in “ethical AI programming.”
LLM Engineer
Large-scale language models (LLMs) are key to today’s AI. In 2017, a research paper titled “attention is all you need” introduced a new concept called the transformer model. This was a breakthrough that opened up AI as we know it today. This gave rise to the concept of large-scale language models that store all the words that the AI has learned and include connections between them based on the context within a sentence or paragraph. LLMs are complex to build and understand, but with the right background you can learn how LLMs work and become an LLM engineer.
Being an LLM engineer doesn’t mean writing code like this: Purpose LLM. This means writing the code to build and operate the LLM itself. These people work for companies like OpenAI, which built ChatGPT. But there are other companies out there, and there are certainly a lot of AI startups. These companies need LLM engineers to write libraries of code to work with LLMs. Where do you start?
First of all, you need to be strong in mathematics. If you think back to learning trigonometric functions in high school, you’ll remember learning how to determine the sine and cosine of a flat (two-dimensional) angle on a page. LLM uses trig to determine how similar human words are to each other. However, trigonometric calculations are performed across hundreds of dimensions, not just two dimensions. Additionally, your mathematical knowledge should extend to advanced calculus, including partial derivatives. A deep knowledge of probability is also required.
The math required is certainly complex, but fortunately, if you studied math, computer science, or physics in college, you’ve probably already learned much of the math you need. You don’t necessarily need to have a master’s degree or a Ph.D. in mathematics. Additionally, you should be familiar with various programming languages such as Java, Python, and C++.
machine learning engineer
Machine learning has been around long before the Transformer model mentioned above. Amazon has done a lot of research into machine learning so that it can, for example, make product suggestions based on your purchases and interactions with your website.
ML engineers need to know how to collect and analyze vast amounts of data, how to “clean” the data (meaning find and remove errors in the data), build large models from it (such as LLM, but could also be other types of models), and how to properly train the models. From there, you need to know how to extract new information from the model. For Amazon, this means extracting new information and generating suggestions based on Amazon’s vast catalog of products.
These models are likely to be huge and require cloud-based platforms such as AWS or Google. (In fact, Amazon Web Services was originally built internally by Amazon to support Amazon’s massive data efforts, and eventually realized that other users could also benefit from using it.)
ML engineering isn’t just limited to product recommendations. One example is healthcare. Large-scale models can be used to predict the likelihood that a patient will develop a particular disease, or to develop protocols to help prevent the disease through specific lifestyle changes or medications. Another example is the supply chain, where machine learning systems can predict demand for a particular product and order the product or the raw materials to build the product.
How do I become an ML engineer? A common way is to start by learning data analysis and data science. Then plan to take a math course similar to the one I described for LLM Engineers. Plan to learn programming in Java, C++, and Python.
AI product manager
AI companies like OpenAI operate very differently than traditional software companies. Their main purpose is to research and develop new ideas in AI. Building software is often a secondary role.
This adds a unique wrinkle to software development, as the job of a researcher is very different from that of a software developer. For example, timelines become more flexible and less predictable. Researchers often conduct many experiments without a clear goal or endpoint in sight. They may need to take several steps back. Multiple false starts can occur. This significantly complicates the management of products based on proprietary research. Also, keep in mind that in many cases, software released by research companies is likely to be a by-product of the research, such as tools for testing and validating the research.
For example, ChatGPT has grown into a premium product for OpenAI, but it was originally a proof of concept for research. In other words, research is often the actual product for such companies, and the tools they create are secondary to the research. This means that managing the creation of such software requires more patience, and the need to recognize that the tools take much more time to build and may need to be completely scrapped if the research changes.
Becoming an AI product manager requires all the skills needed for traditional product management, plus an understanding of how research organizations work. In that case, a basic understanding of how AI, including LLM, works will be helpful.
software developer
While this is by no means a new field, today’s software developers are increasingly being asked to integrate AI into their applications. Something like this is already happening in the app you’re using. For example, Google is rolling out AI assistance on its search pages. To achieve this, developers need to know how to work with AI software libraries.
In this particular case, you don’t need to be an expert in things like building an LLM. Instead, you’ll want to learn as much as you can about LLM and machine learning, and how to use libraries that abstract away the difficult tasks of working with such AI systems. That way, you can focus your time on building your assigned apps and integrating AI without having to become an LLM engineer.
For example, if you write code in Python, you might want to check out ChromaDB. Many popular languages now have similar frameworks. In these cases, you don’t need to know how LLM works, you just need to know how to use it and make it work.
AI quality assurance tester
How on earth do you test an AI tool? Chatbots like ChatGPT are especially difficult to test. In normal software development, when you write a function, you expect that function to give consistent results if you supply it with consistent data. Not so with AI. When we ask questions such as “What is the nature of the universe?” When incorporated into a generative AI tool, it is not possible to verify whether it provides the expected and consistent response.
Traditional QA testing skills are still important when testing chatbots. For example, LLM uses formulas based on trigonometry to determine whether certain words within a given context are similar. The results of these calculations are actually consistent and predictable when using manual QA processes. This is where QA skills come into play in the context of AI software.
This means not only learning everything you can about software testing, but also understanding the big picture of how AI tools work and how parts of AI, such as the LLM, work.
conclusion
AI has created a new industry that will grow for many years to come. The careers introduced here are just a few. Many of the new careers are updates to older careers, such as software developer. When a new AI company is founded, it will need a variety of positions, including customer support, sales representatives, managers, finance personnel, chief technology officers, and CEOs. And all of these roles require at least some understanding of how AI works.
